Narada Chakshu Hackathon Presentation for Territorial Army cyber challenege

Narada Chakshu uses multiple techniques to identify fake videos by tracking eye blinks, analyzing audio-visual sync for lip movement accuracy, detecting skin tone changes via photoplethysmography, checking unnatural head poses, and applying Fourier Transform to uncover AI-generated artifacts.

1. ‘NARADA CHAKSHU’ : A DEEPFAKE DETECTION SOLUTION

2. Project Modules Uses 3D head pose estimation to detect unnatural head movements, and applies Fourier Transform to analyze high-frequency textures, identifying AI-generated artifacts. Detects subtle skin tone variations caused by blood flow, analyzing micro color changes that deepfakes generally fail to replicate. Analyzes lip movements and synchronizes them with the spoken audio to detect mismatches between the two, revealing potential manipulation. Tracks facial landmarks to monitor natural eye blinks and detect anomalies in blink patterns that are often missed in deepfakes.

3. TECH STACK/DEPENDENCIES

4. NEED Combat Misinformation : It's crucial to identify manipulated media to prevent the spread of false information that can mislead the public. Support Legal and Regulatory Compliance : As regulations surrounding media authenticity evolve, having reliable detection tools ensures compliance with legal standards, protecting organizations from potential liabilities. Enhance Technology Research : Developing deepfake detection systems contributes to ongoing research in AI and machine learning, improving understanding of both deepfake creation and detection methodologies. Promote Responsible AI Use : By providing tools to identify deepfakes, we encourage ethical usage of AI technologies, fostering a culture of responsibility and accountability among creators and consumers of digital content. 1 2 3 4 5 6 Protect Privacy & Security : Deepfakes can be used maliciously to create fake identities or damaging content, risking personal privacy and security. A detection system helps safeguard individuals from such threats. Maintain Media Integrity : Ensuring the authenticity of videos is vital for journalism, social media, and other platforms. Deepfake detection helps uphold trust in visual media by distinguishing real content from forged material.

5. STEP 1 Data Collection STEP 4 Audio-Visual Sync Analysis STEP 5 Facial Landmark Detection STEP 2 Preprocessing STEP 3 Pulse Detection STEP 6 Final Detection Matching WORKFLOW

6. Project Submission by Adityavardhan Jain